Artificial intelligence (AI)

AI CHEATSHEET 101 (SAVE THIS )

You’re Using Claude Code Like a Chat Box. It Can Run Itself While You Walk Away.

***YOU CAN READ THIS HERE FOR FREE :)***

AI CHEATSHEET 101

Most people type into Claude Code like it’s a fancy chatbot. A handful of two-word commands turn it into something that works toward a goal on its own, that you can steer from your phone, and that you can rewind like a video game.

Here is the quiet truth about Claude Code. Most people are using maybe five percent of it.

They open it, type a request, wait, type another, wait again. Basically a smarter chat box that happens to touch their files. And that is fine, it works. But under the hood there is a whole layer of little slash commands, the ones that start with a forward slash, that completely change what this thing is. Not what you type into it. How you work with it.

I want to show you the few that actually matter. Not all of them, just the handful that made me go oh, I have been doing this the hard way this whole time. This will take you about three minutes and it might genuinely change your week.

The one that lets you walk away

Start with the big one. There is a command called /goal.

You give it a finish line, something like “get all the tests passing” or “migrate this old code to the new setup,” and then you let go. Claude keeps working toward that goal turn after turn, on its own, figuring out the path itself, with a little progress bar showing you where it is. You are not standing there feeding it the next instruction every thirty seconds. You define what done looks like, and it goes and gets there.

That single shift, from babysitting every step to setting a destination and stepping back, is the difference between using a tool and actually delegating to one.

The one that fits in your pocket

Now pair that with the move that genuinely surprised me. It is called /remote-control, or just /rc for short.

You load up your big task on your computer, hit /rc, and then you can steer the whole session from your phone. Walk to the kitchen. Go for a walk. Sit on the couch. And from your phone you nudge it, answer its questions, point it in a new direction, while the actual work grinds away on your machine.

Put those two together and the picture gets a little wild. You set a goal, switch to remote control, and your computer is quietly working toward a finish line while you are out living your life, checking in from your pocket now and then like a manager glancing at a text. That is not a chat box. That is a tiny employee.

The one that saves you from yourself

Here is the one that takes the fear out of letting Claude loose. It is called /rewind.

Ever watched it confidently head down completely the wrong path, making a mess across your files, and felt that little spike of panic? With /rewind, you just roll the whole thing back. Not only the conversation, but the actual code and file changes too, back to an earlier checkpoint, like hitting undo on the entire situation. You went down a dead end, so you rewind and try again, clean. No damage, no junk left behind.

Once you know you can always rewind, you stop being so cautious, and you let it move faster.

The two tiny ones you should use constantly

A couple of small habits punch way above their weight.

The first is /clear. The moment you finish one task and move to something unrelated, clear the conversation. A fresh, empty slate keeps Claude sharp and stops it dragging old, irrelevant junk into your new problem, which is what makes its answers go fuzzy. If your next request has nothing to do with the last one, clear first.

The second is /btw, as in “by the way.” It lets you fire off a quick side question while a long task is still running, without stopping it. The big job keeps going in the background while you get your little answer on the side. Small thing, oddly satisfying, weirdly useful.

The whole thing in one smooth flow

Let me hand you the workflow that ties it together, because this is the part worth stealing.

You start a session and immediately name it, so you can find it later. You load up your context and set your goal, so it can run on its own. You flip to remote control and steer from your phone while you go do something else. When you come back, you quickly review the changes it made and check what it cost you. And when the job is done, you clear it out, ready to jump back in later if you need to.

That is the rhythm the people getting real leverage out of this tool have quietly settled into. Set a destination, let it run, steer from a distance, review, reset. The ones still typing one prompt at a time and waiting are doing honest work with a Ferrari stuck in first gear.

You do not need to memorise two dozen commands. You need these few, and the willingness to stop treating Claude Code like a chatbot. It was never meant to be one.

So tell me, were you stuck in chat box mode too, or is there a command you swear by that I left off? Drop it in the comments, I am always stealing better workflows.

Give Me A Follow

If this leveled up how you use Claude Code, do me a quick favour. Clap, save it so the commands are there when you need them, and leave a comment. On Medium those three little taps are what tell the algorithm a post is worth showing to more people, so even a quick tap genuinely helps this reach the next person stuck typing one prompt at a time. It would mean a lot. Thank you for reading.